Evaluation of an indirect serum ELISA and a bacteriological faecal culture test for diagnosis of Salmonella serotype Dublin in cattle using latent class models.

Abstract

To evaluate a conventional bacteriological test based on faecal culture and an indirect serum ELISA for detection of S. Dublin infected cattle. To compare the predictive values of the two tests in relation to the prevalence. A total of 4531 paired samples from cattle in 29 dairy herds were analysed for presence of S. Dublin bacteria in faeces and immunoglobulins directed against S. Dublin lipopolysaccharide in an indirect serum ELISA. Sensitivity and specificity were estimated at two ELISA cut-off values using a validation method based on latent class models, which presumably provides less biased results than traditional validation methods. Stratification of data into three age groups gave significantly better estimates of test performance of the ELISA. Receiver operating characteristic (ROC) curves were constructed for comparison of overall performance of the ELISA between the three age groups. The sensitivity of the faecal culture test was low (6-14%). ELISA appeared to have a higher validity for animals aged 100-299 days of age than older or younger animals. Overall, the negative predictive value of the ELISA was 2-10 times higher than for the faecal culture test at realistic prevalence of infection in the test population. The diagnostic sensitivity of the faecal culture test for detection of S. Dublin is poor, the specificity is 1. The superior sensitivity and negative predictive value of the serum ELISA makes this test preferable to faecal culture as an initial screening test and for certification of herds not infected with S. Dublin. A quantitative estimate of the sensitivity of a faecal culture test for S. Dublin in a general population was provided. ELISA was shown to be an appropriate alternative diagnostic test. Preferably, samples from animals aged 100-299 days of age should be used as these give the best overall performance of the ELISA. Plots of ROC curves and predictive values in relation to prevalence facilitates optimisation of the ELISA cut-off value.